INDEX

Explanations

connecting concepts

The neuron strongly activates on the coordinating conjunction “and.”

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 både

-3.86

ò

-3.27

辚

-2.98

先是

-2.94

 alors

-2.92

/",

-2.92

/"

-2.88

 også

-2.88

 zarówno

-2.78

 jedoch

-2.77

POSITIVE LOGITS

 sebagainya

3.55

 یا

3.38

 possibly

3.25

or

3.23

This

3.20

 then

3.09

 whatnot

2.98

2.94

 three

2.91

After

2.83

Activations Density 1.660%